Running exercise strengthens the intervertebral disc

Abstract

There is currently no evidence that the intervertebral discs (IVDs) can respond positively to exercise in humans. Some authors have argued that IVD metabolism in humans is too slow to respond anabolically to exercise within the human lifespan. Here we show that chronic running exercise in men and women is associated with better IVD composition (hydration and proteoglycan content) and with IVD hypertrophy. Via quantitative assessment of physical activity we further find that accelerations at fast walking and slow running (2 m/s), but not high-impact tasks, lower intensity walking or static positions, correlated to positive IVD characteristics. These findings represent the first evidence in humans that exercise can be beneficial for the IVD and provide support for the notion that specific exercise protocols may improve IVD material properties in the spine. We anticipate that our findings will be a starting point to better define exercise protocols and physical activity profiles for IVD anabolism in humans.

Full text

1
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
www.nature.com/scientificreports
Running exercise strengthens the
intervertebral disc
Daniel L. Belavý1, Matthew J. Quittner1, Nicola Ridgers1, Yuan Ling2, David Connell2,3 &
Timo Rantalainen1
There is currently no evidence that the intervertebral discs (IVDs) can respond positively to exercise in
humans. Some authors have argued that IVD metabolism in humans is too slow to respond anabolically
to exercise within the human lifespan. Here we show that chronic running exercise in men and
women is associated with better IVD composition (hydration and proteoglycan content) and with IVD
hypertrophy. Via quantitative assessment of physical activity we further nd that accelerations at fast
walking and slow running (2 m/s), but not high-impact tasks, lower intensity walking or static positions,
correlated to positive IVD characteristics. These ndings represent the rst evidence in humans that
exercise can be benecial for the IVD and provide support for the notion that specic exercise protocols
may improve IVD material properties in the spine. We anticipate that our ndings will be a starting point
to better dene exercise protocols and physical activity proles for IVD anabolism in humans.
We expect that tissues will adapt to loads placed upon them. Wol1 described the theory of bone adaptation to
loading. In the intervening years, evidence2 has been obtained as to which loading protocols are benecial (oste-
ogenic) for bone. For the intervertebral disc (IVD), little is known about what loading protocols are benecial
for IVD tissue and cause anabolism in humans. We have good knowledge of loading types that are more likely to
damage lumbar IVD tissue in humans, such as exion of the spine with compression3, torsion4 or to damage to
the vertebral end-plate via axial compression3 with subsequent IVD degeneration5. Whilst this information can
inform what activities people should avoid to preserve IVD integrity, it does not inform us on exercise or habitual
physical activity to “strengthen” the IVD. Furthermore, data6 on turnover rates in the IVD, lead to the assumption
that positive adaptation in the mature IVD is unlikely to occur during the normal human lifespan.
Currently we rely on data from animal, IVD cell and IVD tissue models to suggest what kind of loading might
be benecial for the human lumbar IVD. ese models suggest7 that a “likely anabolic loading window” for the
IVD exists: dynamic loading of 0.2–0.8 MPa, generating intra-discal pressures of approximately 0.3–1.2 MPa, at
0.1 to 1 Hz for approximately eight hours a day. Given human data on intra-discal pressures in dierent activities8,
this could9 be extrapolated to suggest that walking or running exercise is likely anabolic for the IVD. Quadrupedal
treadmill running exercise in rodents10,11 can have a positive impact on the rodent IVD. However, directly apply-
ing loading thresholds and protocols from animal models to humans is problematic12 and there is no evidence
yet9 of a benecial eect of exercise on the IVD in humans. We aimed to determine whether benecial eects on
the IVD of exercise can be seen in humans and what loading patterns this might entail.
Our hypothesis was that people who perform regular upright running activity will show better IVD tissue
quality, as shown by higher T2-times13 in their lumbar IVDs, than people who are healthy with no history of spi-
nal disease, but otherwise not physically active. We also hypothesised that there would be a dose-response eect
of dierent volumes of running. Furthermore, to better understand what types of physical activity are likely ben-
ecial for IVD, we explored the relationship between habitual physical activity, as measured by objective accel-
erometry, and IVD characteristics. To reduce the confounding inuence of normal aging on our ndings, and
given evidence9 that IVD maturation is still in process in the third decade of life, we included women and men
aged 25–35 years. It is also not clear how long is required before the IVD might show a measurable adaptation
to exercise, and we therefore recruited only people with a minimum of 5 years history at their current physical
activity level: either no sport (referents), 20–40 km per week running (joggers), or 50 + km per week running
(long-distance runners).
1Deakin University, School of Exercise and Nutrition Sciences, Institute for Physical Activity and Nutrition, 221
Burwood Highway, Burwood, Victoria, 3125, Australia. 2Imaging at Olympic Park, 60 Olympic Boulevard, Melbourne,
Victoria, 3004, Australia. 3Monash University, Wellington Road, Clayton, Victoria, 3168, Australia. Correspondence
and requests for materials should be addressed to D.L.B. (email: belavy@gmail.com)
Received: 24 November 2016
accepted: 07 March 2017
Published: 19 April 2017
OPEN

www.nature.com/scientificreports/
2
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
Results
Long-distance runners (Table1) and joggers showed signicantly higher (+ 11.4% and + 9.2% respectively) lum-
bar IVD T2-times than the non-athletic individuals (Fig.1). is eect was also present at all individual vertebral
levels T11/T12 to L5/S1 (Fig.2; upper panel). e eect of running on T2-time was strongest in the IVD nucleus
(Fig.3; + 11% in joggers and + 15% in long-distance runners in the central nuclear region versus + 5% to + 6%
respectively in the anterior annulus and + 5% and + 9% respectively in the posterior annulus). e height of the
IVD relative to that of the vertebral body, an indicator of IVD hypertrophy, was greater in the long-distance run-
ners (Fig.1). When examining individual vertebral levels (Fig.2; lower panel), this eect was present at the lower
No spor t
Running
(20–40 km)
Running
(50 k + )
Number of males (of
total N) 11 of 24 13 of 30 11 of 25
Body mass (kg) 73.9(17.8) 68.2(11.1) 63.5(10.2)*
Age (yrs) 29.3(3.7) 30.2(3.2) 30.1(3.9)
Height (cm) 173.2(8.7) 173.6(9.7) 170.3(9.2)
Weekday sitting time
(hrs) 9.7(2.1) 6.3(2.8)‡ 6.4(2.8)‡
Exercise
participation (yrs) - 8.8(4.2) 7.6(3.9)
Exercise
participation (hrs/
wk)
- 4.9(2.2) 8.6(4.3)
Exercise distance
(km/wk) - 28.0(6.7) 66.6(19.5)
IVD volume (cm3) 9.5(2.3) 10.1(3.4) 10.1(3.3)
IVD average area
(mm2)250.2(41.2) 263.6(60.8) 263.0(58.9)
IVD height (mm) 7.1(0.7) 7.3(1.1) 7.5(1.0)
IVD anteroposterior
width (mm) 25.6(2.4) 26.4(3.5) 25.9(3.4)
Intervertebral
distance (mm) 34.2(2.0) 34.1(2.9) 34.0(2.8)
Prrmann grade 2.3(0.39) 2.2(0.39) 2.1(0.39)
Erector spinae size
(cm2)14.3(3.3) 14.4(5.0) 14.1(4.8)
Lumbar multidus
size (cm2)4.7(1.1) 4.4(1.6) 4.1(1.6)
Psoas size (cm2) 9.1(2.8) 10.1(4.2) 9.6(4.0)
Quadratus
lumborum size
(cm2)
3.5(1.1) 3.4(1.7) 3.1(1.6)
Table 1. Participant characteristics, intervertebral disc characteristics and lumbar muscle morphology.
Values of continuous variables are mean(SD). *p < 0.05; †p < 0.01; ‡p < 0.001 and indicate signicance of
dierence to the non-sporting group. e proportion of females:males did not dier across groups (χ
2 = 0.01,
p = 0.99). ree participants were of Asian descent (one female 20–40 km runner, one male 20–40 km runner,
one male 50 + km runner) and the remaining participants were Caucasian. Prrmann grade was averaged from
all lumbar discs. Areas of erector spinae, multidus, psoas and quadratus lumborum muscles were averaged
from le and right sides of the body and all lumbar levels.
Figure 1. Runners have more hydrated (le) and hypertrophied (right) lumbar IVDs. Values are
mean(SD) averaged across all lumbar discs. Le panel: Higher T2-times indicate13 better IVD hydration and
glycosaminoglycan content. Right panel: IVD height relative to vertebral body height. *p < 0.05; †p < 0.01 and
indicate signicance of dierence to the non-sporting group.

www.nature.com/scientificreports/
3
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
lumbar vertebral levels L3/L4 to L5/S1. Lumbar muscle size did not dier between groups (Table1). e eect of
running exercise was consistent for both genders, with the gender × group interaction not reaching signicance
(p > 0.18). e long-distance group typically showed greater dierences in IVD parameters to the non-sporting
group than the jogging group. However, there were no statistically signicant dierences between the two run-
ning groups.
Total physical activity levels, as measured by objective accelerometry, were not related to IVD characteristics
(Fig.4). Rather, IVD nucleus T2-time was most strongly associated with accelerations in the range 0.44 and 0.59 g
mean amplitude deviation (MAD; Fig.5). Additional accelerometry data collected under dierent exercise con-
ditions in ten individuals (SupplementaryTable1) showed that ambulation at 2 m/s fell inside this 0.44 and 0.59 g
MAD range. Walking at 1.5 m/s or slower fell below this range and running at 2.5 m/s or faster and jumping were
above this range.
Discussion
Before the rst interventional studies to dene exercise regimes for improving bone characteristics were per-
formed in the 1990s, an important step was the nding of dierences in bone density between dierent athletic
populations14. is built on prior animal studies and helped to show that exercise may well result in anabolic
adaptation of bone. In this vein, the current study builds on prior work in animal, cell and tissue explant models7
and provides the rst ever cross-sectional evidence in humans that exercise may well favourably impact the IVD.
Our main nding was that long-distance runners and joggers showed better hydration and glycosaminogly-
can levels (higher lumbar IVD T2-times13) than the non-athletic individuals. is is consistent with ndings of
an anabolic response in the IVD in quadrupedal animals10,11 to running. e nding is also consistent with the
notion, developed on the basis of cell, tissue explant and animal models7,15, of a “likely anabolic window” for
IVD loading. e eect of running in humans on IVD composition was most obvious in the IVD nucleus where
intra-discal pressure increases with applied axial load are constrained by the ring-formed annular bres16.
Beyond the compositional dierences, there was evidence of IVD hypertrophy in the long-distance runners.
e height of the IVD relative to that of the vertebral body, which serves as an internal control for body size, was
greater in the long-distance runners. is extends on ndings17 from monozygotic twins that IVDs were margin-
ally, but not signicantly, larger in those twins that were at least 8 kg heavier than their twin pair and presumably
experienced greater habitual spinal loading. Hypertrophy of the IVD may well be an adaptation to habitual load-
ing in runners. Similar to hypertrophic responses seen in muscle due to resistance training18, this suggests that
tissue adaptation will occur in the IVD with exercise. Overall our ndings provide support for the hypothesis that
an adaptive, anabolic and hypertrophic response is possible in the human IVD with exercise.
e current study also provides some guidance on what kinds of loading protocols may be better for the IVD.
To understand what kinds of physical activity in humans might be the drivers of an anabolic response in the IVD,
we examined the physical activity patterns of our collective via objective accelerometry. Total physical activity
levels were not related to positive adaptations in the IVD, rather accelerations in a specic range. e strongest
association to higher IVD T2-times were seen between 0.44 and 0.59 g mean amplitude deviation. is ts with
the idea from animal, tissue and cell models7 of a “likely anabolic window” for the IVD. To better understand
what kinds of activities generate these acceleration magnitudes, we collected additional accelerometry data under
dierent conditions. Walking or slow running at 2 m/s fell inside this range with slower walking falling below this
range. Fast running and high-impact jumping activities were above this range. is is in line with the notion that
high-impact loading is considered7,19 to be detrimental to the IVD and vertebral end-plate. Dynamic IVD load-
ing of 0.2–0.8 MPa, generating intra-discal pressures of approximately 0.3–1.2 MPa, is thought7 to be an optimal
loading magnitude for the IVD. Based upon data on in vivo intradiscal pressures8, activities such as walking and
running, but not liing a 20 kg load or lying down, fall into this loading magnitude window. is corresponds
well to our observations here of the impact of exercise in athletes. We also noted that sedentary activities were
unrelated to IVD characteristics. In light of prior work, the results of our study suggest that, in comparison to
other locomotor activity, fast walking or slow running may provide the strongest anabolic stimulus for adaptation
in the IVD in humans.
Whilst the long-distance running group showed consistently better IVD properties than the jogging group,
there were no statistically signicant dierences between the two. Also, there was no relationship between the
IVD characteristics and physical activity in the 0.7 to 0.9 g MAD range where the physical activity associated with
running was most evident (SupplementaryFigure1). is indicates a ceiling eect of exercise for the IVD was
approached for both volume of upright axial spine loading and intensity. A ceiling eect in relation to exercise
has been observed for muscle hypertrophy20 and bone adaptation to exercise21. It is possible that high volume or
intensity running is not required for a benecial adaptation in the IVD.
In the wider population, it is the lower lumbar IVDs that are most commonly aected by degeneration22.
Furthermore, repetitive loading of the spine is considered23 to be a contributory factor to the development of IVD
degeneration. Despite repetitive loading of the spine during running, the exercise groups of the current study
did not show any detrimental eects at these lower lumbar segments. In contrast, the long-distance runners and
joggers showed evidence of better IVD hydration and glycosaminoglycan content in the lower lumbar spine than
those that did not perform sport. Furthermore, the evidence for IVD hypertrophy subsequent to habitual run-
ning was strongest at the lower lumbar levels. Our data show that repetitive axial loading of the spine under body
weight during running in otherwise healthy people may well be benecial for the lower lumbar IVDs.
It is important to consider some of the limitations of the current study. We performed a cross-sectional study
as a rst step to examine whether certain types of exercise might be benecial for the IVD in humans. In this
design it is not possible to completely rule out other confounding factors. We showed, for example, that lumbar
muscle size was similar in all groups and this indicates that muscle adaptation per se is not the likely cause of

www.nature.com/scientificreports/
4
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
Figure 3. e impact of running on the disc is strongest in the nucleus. Top: 3D plot of mean T2-time across
entire IVD volume. Bottom: At the mid-line (sagittal) portion of the IVD the impact of running can be seen to
be greatest in the central, nuclear, portion of the IVD. *p < 0.05; †p < 0.01 versus non-sporting group. Greater
T2-times indicate13 better IVD hydration and glycosaminoglycan content.
Figure 2. Eect of running is also present at lowest lumbar vertebral levels: T2-time (top) and IVD height
relative to vertebral body height (bottom). Values are mean(SD) at each vertebral level. *p < 0.05; †p < 0.01;
‡p < 0.001 and indicate signicance of dierence to the non-sporting group. VB: vertebral body.

www.nature.com/scientificreports/
5
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
dierences in IVD characteristics in our populations. However, we cannot rule out other factors such as dif-
ferences in muscle function, dierences in nutrition, systemic hormonal (e.g. growth factors, cytokines, stress
hormones) dierences, or other indirect eects. To denitively determine that the mechanical loading from spe-
cic exercise forms result in positive adaptations in the IVD, to determine mechanisms of action, and delineate
exercise guidelines for strengthening the IVD, randomised controlled exercise trials are necessary.
at an optimal loading level and pattern for the IVD exists makes sense from a tissue homeostatis perspec-
tive. Given that tissues such as bone21 and muscle18 have optimal loading conditions for an anabolic response, for
the IVD we should expect it to be no dierent. Our ndings support the perspective that, when loaded appropri-
ately, anabolic IVD adaptation can occur in humans and that this can occur on time frame for it to be meaningful
within the human lifespan.
Evidence that the IVD will respond anabolically in humans to certain types of loading may have public health
implications. Spinal pain consistently presents one of the greatest costs to developed societies for disability and
lost productivity, including when increased death rates in other diseases are accounted for24. IVD degeneration
and herniation is one important contributing factor to spinal pain. Similar to understanding the impact of specic
exercise in other disease constellations, for example in type II diabetes25, knowing that the IVD can respond to
certain kinds of loading, and understanding what kinds of loading are optimal, will result in better exercise guide-
lines for the prevention and management of spinal pain.
Figure 4. Total physical activity levels were unrelated to IVD characteristics. Empty symbols = males, lled
symbols = females. Individual values for each subject shown. ere was no correlation between total physical
activity levels and T2-time in the intervertebral disc (IVD) nuclear region.
Figure 5. Specic loading levels impact the IVD nucleus T2-time. Values are mean(95% CI) correlation
between IVD nucleus T2-time and count data in each mean amplitude deviation bin. e strongest association
to higher IVD T2-times were seen between 0.44 and 0.59 g. is corresponds to fast walking and slow jogging
(see Results and SupplementaryTable1).

www.nature.com/scientificreports/
6
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
Methods
Ethical approval and subjects. e study was approved by the Deakin University Faculty of Health,
Medicine, Nursing & Behavioural Sciences human ethics advisory group. All subjects gave their informed writ-
ten consent prior to participation in the study. All methods were performed in accordance with the relevant
guidelines and regulations. To reduce the impact of normal ageing on the ndings, only individuals aged 25 to
35 years of age were included. Exclusion criteria included current spinal pain, history of spinal surgery, history of
traumatic injury to the spine, known scoliosis for which prior medical consultation was sought, current or prior
smoker, known claustrophobia and possible pregnancy. We recruited three groups of people with distinct loading
histories: joggers (20–40 km per week), long-distance runners (50+ km per week) and non-sporting referents.
To be included in the jogging group, subjects needed to have been running 20–40 km per week for a minimum
of the last 5 years and perform no other sport or exercise type more than once per week. Long-distance runners
were required to have performed at least 50 km per week for a minimum of the last 5 years and, with the exception
of resistance exercise for muscle hypertrophy (muscle hypertrophy training is a common training component
of long-distance runners), no other sport or exercise type more than once per week. Included in the “no-sport”
group were individuals who performed no regular sport or exercise in the last ve years, currently performed
less than 150 minutes of moderate activity (dened as activity that “causes an individual to breathe harder than
normal”) per week26, and walked less than 15 min to or from their place of work. A total of 79 participants were
included in the study (Table1).
Testing and scanning protocol. Participants were instructed not to perform any exercise on the day of
their scan. Due to normal diurnal variation in IVD water content, all testing was performed aer midday. Upon
arriving at the radiology facility, participants were required to sit for a minimum of 20 minutes prior to entering
the scanner with participants sitting for a mean(SD) of 44(16) minutes which did not dier between groups
(p > 0.19). During this time participants completed questionnaires detailing their gender, type of physical activity,
body height, and average sitting duration Monday to Friday. e runners also reported distance run per week,
time run per week and number of years of participation.
To quantify IVD T2-time and morphology, a spin-echo multi-echo sequences on a 3T Phillips Ingenia scanner
(Amsterdam, Netherlands) was used with spinal coils to collect images at 8 echo times (15.75, 36.75, 57.75, 78.75,
99.75, 120.75, 141.75 and 162.75 ms) from 13 sagittal anatomical slices each (thickness 3 mm; interslice distance:
1.5 mm; repetition time: 2000 ms, eld of view: 281 × 281 mm, image resolution: 0.366 mm per pixel) encom-
passing the entire lower spine from le to right. For radiological categorisation of IVD degeneration (Prrmann
grade), a sagittal plane T2-weighted sequence (15 slices, slice thickness: 3 mm, interslice distance: 1.5 mm, repeti-
tion time: 2600 ms, echo time: 70 ms, eld of view: 357 mm × 357 mm, resolution: 0.532 mm per pixel) was taken.
To quantify muscle morphology, a paraxial T1-weighted scan (repetition time: 800 ms, echo time: 9 ms, slice
thickness: 4 mm, interslice distance: 2 mm, eld of view: 258.68 × 258.68 mm, image resolution: 0.270 mm per
pixel) with ve groups of three slices each positioned at each vertebral body L1 to L5 and oriented to the vertebral
end-plates was performed. Data were exported for further oine processing.
Aer scanning, subjects were given a hip-mounted ActiGraph model GT3X+ (Pensacola, FL). Participants
were instructed to wear the ActiGraph during all waking hours except during water-based activities (e.g. swim-
ming and bathing) for eight consecutive days. Acceleration data were collected at 100 Hz with a ±6 g range and
12 bit analog to digital conversion.
Oine image processing and analysis. To ensure blinding of the examiner to oine image measure-
ments, each subject was assigned a random numeric code (obtained from www.random.org). A radiologist deter-
mined the Prrmann grade of each lumbar IVD (Table1) on sagittal T2-weighted images. Seven individuals had
a supernumerary lumbar vertebral segment and the additional IVD (designated L6/S1) in these subjects was not
included in analyses.
ImageJ 1.38x (http://rsb.info.nih.gov/ij/) was used to perform all quantitative MRI measures. In the sagittal
spin-echo multi-echo images every IVD from T11/T12 to L5/S1 was measured. Aer segmenting the IVD, a
custom written ImageJ plugin (“ROI Analyzer”; https://github.com/tjrantal/RoiAnalyzer) was used to rotate the
region of interest to the horizontal and measure area, height, width and signal intensity of the IVD in its entirety
as well as in ve subregions from the anterior to posterior aspect of the disc. T2-time was calculated via linear t
to the natural logarithm of the image intensity in each of the eight MR echos. Similar parameters were generated
for each of the ve disc subregions and interpolated across the width of the IVD to generate the 3D plots pre-
sented in Fig.3. Volume of each IVD in each subject was calculated by linear interpolation of the area data from
all slices. e image number where each spinous process was best visible was noted. e vertebral body was also
segmented and vertebral body height measured in a similar fashion in order to calculate the ratio of IVD height to
vertebral body height, as a normalised indicator of IVD hypertrophy. e data from each lumbar IVD were also
averaged. With the exception of IVD volume (Table1), data averaged from the three images around the spinous
process were used in analysis.
In each of the paraxial T1-weighted images, area of the lumbar multidus, erector spinae, psoas and quadratus
lumborum were measured bilaterally from L1 to L5 as in prior work27. e muscle area averaged from all lumbar
levels was used in analysis.
3D accelerometry analysis. Once participants returned the Actigraph, raw data were downloaded from
the device and analysed with a custom-written Matlab script (R2015b, Mathworks, Inc., Natick, MA, USA).
Resultant acceleration was calculated from the 3-dimensional data, and used in all further analyses. No smooth-
ing was applied on the recorded signal.

www.nature.com/scientificreports/
7
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
Mean amplitude deviation ( =
∑−
=
MA
D
xx
n
i
n
i
1, where n = number of data points in an epoch, i = data point
index, x = resultant acceleration,
x
= mean of the epoch, and vertical bars signify taking the absolute value)28 was
calculated from the resultant acceleration in non-overlapping 5 s epochs in 24 h segments starting at 6 am local
time for all of the recorded days. e 5 s epoch mean MADs were divided into 98 logarithmically equidistant bins
(= histogram) from 0 g to 2.5 g. Non-wear time was dened as any hours with standard deviation less than 0.024 g,
and any days with less than 10 h of total wear-time were excluded from analysis. Data from individuals with ≥ 3
days of sucient wear-time were included. 70 individuals completed the Actigraph data collection and were
included in the analysis. Two individuals from the no-sport group, six from the jogging group and one from the
long-distance running group did not complete this data collection. On average, 7.6 days of complete data were
available per participant (similar across all three groups, p = 0.9). e mean of all included days is reported.
Additional sub-study: which activities generate what accelerations? To relate MAD to physical
activities, 10 individuals performed a graded treadmill test at speeds of 0.5 m/s, 1 m/s, 1.5 m/s, 2 m/s, 2.5 m/s,
3.0 m/s and 3.5 m/s whilst wearing a hip-mounted ActiGraph. At each speed, 70 seconds of data were collected
and the data from the 11 second onwards was used for further analysis. Participants also performed 10 consec-
utive jumps maintaining the knee and hip at near full extension (similar to hopping on one leg, but bilaterally).
Statistical analyses. An alpha-level of 0.05 was taken for statistical signicance. For continuous variables,
T-tests were performed comparing the long-distance and jogging groups to the non-sport referent group. Primary
analysis evaluated IVD T2-time (composition measure) and IVD height to vertebral body height ratio (hypertrophy
measure) averaged across all lumbar levels. Data from individual vertebral levels were also evaluated. Dierence in
response between males and females was evaluated via two-way analysis of variance for ‘group’ and ‘gender’. For com-
paring physical activity to IVD characteristics, the correlation and 95% condence interval was calculated between
number of counts for each subject in the MAD bin and the average lumbar IVD T2-time in the nucleus (central
subregion). e mean and 95% condence interval of the MADs in each of the physical activities of the sub-study
were calculated. e “R” statistical environment (version 2.10.1, www.r-project.org) was used for all analyses.
References
1. Wol, J. e law of bone remodeling (Das Gesetz der Transformation der Knochen). (Springer-Verlag Berlin Heidelberg, 1892).
2. Xu, J., Lombardi, G., Jiao, W. & Ban, G. Eects of Exercise on Bone Status in Female Subjects, from Young Girls to Postmenopausal
Women: An Overview of Systematic eviews and Meta-Analyses. Sports Med. 46, 1165–1182 (2016).
3. Adams, M. A. & Hutton, W. C. Prolapsed intervertebral disc. A hyperexion injury 1981 Volvo Award in Basic Science. Spine 7,
184–191 (1982).
4. Schmidt, H., Heuer, F. & Wile, H.-J. Dependency of disc degeneration on shear and tensile strains between annular ber layers for
complex loads. Med. Eng. Phys. 31, 642–649 (2009).
5. Wang, Y., Videman, T. & Battié, M. C. ISSLS prize winner: Lumbar vertebral endplate lesions: associations with disc degeneration
and bac pain history. Spine 37, 1490–1496 (2012).
6. Chan, S. C. W., Ferguson, S. J. & Gantenbein-itter, B. e eects of dynamic loading on the intervertebral disc. Eur. Spine J. 20,
1796–1812 (2011).
7. Wile, H. J., Neef, P., Caimi, M., Hoogland, T. & Claes, L. E. New in vivo measurements of pressures in the intervertebral disc in daily
life. Spine 24, 755–762 (1999).
8. Belavý, D. L., Albracht, ., Bruggemann, G.-P., Vergroesen, P.-P. A. & van Dieën, J. H. Can Exercise Positively Influence the
Intervertebral Disc? Sports Med. 46, 473–485 (2016).
9. Brisby, H. et al. e eect of running exercise on intervertebral disc extracellular matrix production in a rat model. Spine 35,
1429–1436 (2010).
10. Sasai, N. et al. Physical exercise aects cell proliferation in lumbar intervertebral disc regions in rats. Spine 37, 1440–1447 (2012).
11. Alini, M. et al. Are animal models useful for studying human disc disorders/degeneration? Eur. Spine J. 17, 2–19 (2008).
12. Sivan, S. S. et al. Aggrecan turnover in human intervertebral disc as determined by the racemization of aspartic acid. J Biol Chem
281, 13009–14 (2006).
13. Marinelli, N. L., Haughton, V. M., Munoz, A. & Anderson, P. A. T2 relaxation times of intervertebral disc tissue correlated with water
content and proteoglycan content. Spine 34, 520–4 (2009).
14. Bloc, J. E. et al. Determinants of bone density among athletes engaged in weight-bearing and non-weight-bearing activity. J. Appl.
Physiol. 67, 1100–1105 (1989).
15. Iatridis, J. C., MacLean, J. J., oughley, P. J. & Alini, M. Eects of mechanical loading on intervertebral disc metabolism in vivo. J
Bone Jt. Surg Am 88 Suppl 2, 41–6 (2006).
16. Schmidt, H. et al. Intradiscal pressure, shear strain, and ber strain in the intervertebral disc under combined loading. Spine 32,
748–55 (2007).
17. Videman, T., Gibbons, L. E., aprio, J. & Battié, M. C. Challenging the cumulative injury model: positive eects of greater body mass
on disc degeneration. Spine J. 10, 26–31 (2010).
18. American.College.of.Sports.Medicine. American College of Sports Medicine position stand. Progression models in resistance
training for healthy adults. Med Sci Sports Exerc 41, 687–708 (2009).
19. oaf, . A study of the mechanics of spinal injuries. J. Bone Joint Surg. Br. 42, 810–823 (1960).
20. raemer, W. J. & atamess, N. A. Fundamentals of resistance training: progression and exercise prescription. Med. Sci. Sports Exerc.
36, 674–688 (2004).
21. Turner, C. H. ree rules for bone adaptation to mechanical stimuli. Bone 23, 399–407 (1998).
22. Miller, J. A., Schmatz, C. & Schultz, A. B. Lumbar disc degeneration: correlation with age, sex, and spine level in 600 autopsy
specimens. Spine 13, 173–178 (1988).
23. Adams, M. A. & oughley, P. J. What is intervertebral disc degeneration, and what causes it? Spine 31, 2151–2161 (2006).
24. Murray, C. J. L. et al. Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: a systematic
analysis for the Global Burden of Disease Study 2010. Lancet Lond. Engl. 380, 2197–2223 (2012).
25. Laasonen, D. E. et al. Physical activity in the prevention of type 2 diabetes: the Finnish diabetes prevention study. Diabetes 54,
158–165 (2005).
26. World Health Organization. Global recommendations on physical activity for health. (2010).
27. Belavý, D. L., Armbrecht, G., ichardson, C. A., Felsenberg, D. & Hides, J. A. Muscle atrophy and changes in spinal morphology: is
the lumbar spine vulnerable aer prolonged bed-rest? Spine 36, 137–145 (2011).
28. Vähä-Ypyä, H., Vasanari, T., Husu, P., Suni, J. & Sievänen, H. A universal, accurate intensity-based classication of dierent
physical activities using raw data of accelerometer. Clin. Physiol. Funct. Imaging 35, 64–70 (2015).

www.nature.com/scientificreports/
8
Scientific RepoRts | 7:45975 | DOI: 10.1038/srep45975
Acknowledgements
We thank the subjects for participating in the study. We also thank the sta at Imaging at Olympic Park for their
support in implementing the study. is project was supported by internal institutional funding from Deakin
University School of Exercise and Nutrition Sciences (reference number ‘Belavy 2014–2017′ ).
Author Contributions
D.L.B.: Contributed conception and design of the experiments, image analysis, statistical analysis, interpretation
of the data, draing the article. M.Q.: contributed subject recruitment, data collection, image analysis, approved
manuscript. N.R.: contributed to conception and design of the experiments, analysis of actigraph data, revision of
the manuscript, approved manuscript. Y.L.: contributed radiological assessment of images, approved manuscript.
D.C.: contributed to conception and design of the experiments, support with data collection, interpretation of
the data, approved manuscript. T.R . : contributed to conception and design of the experiments, data analysis,
statistical analysis, draing manuscript, approved manuscript.
Additional Information
Supplementary information accompanies this paper at http://www.nature.com/srep
Competing Interests: e authors declare no competing nancial interests.
How to cite this article: Belavý, D. L. et al. Running exercise strengthens the intervertebral disc. Sci. Rep. 7,
45975; doi: 10.1038/srep45975 (2017).
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional aliations.
is work is licensed under a Creative Commons Attribution 4.0 International License. e images
or other third party material in this article are included in the article’s Creative Commons license,
unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license,
users will need to obtain permission from the license holder to reproduce the material. To view a copy of this
license, visit http://creativecommons.org/licenses/by/4.0/
© e Author(s) 2017